Reflections of low-frequency, broadband airgun signals from ocean water columns have long been collected in modern marine seismic surveys. However, they were barely examined because of their weak amplitude (following from low acoustic impedance contrast in water columns) and the lack of application interests. Nevertheless, in the past decades, a new cross-discipline “seismic oceanography” has developed the interest to use these water-column reflection signals to image the ocean structures. Motivated by seismic oceanography applications, we examined the acoustic multipath structure of marine seismic survey data with a focus on water-column reflections, and developed a two-step matched filtering approach to enhance water-column reflection and suppress the unwanted bubble waves. The approach was applied to process data collected from the Gulf of Mexico and led to an improvement in imaging mesoscale ocean structures when compared with the traditional matched filtering approach. For the specific data we examined, the results reveal a 11.3-dB improvement of signal-to-noise ratio by removing the noise and a 8-dB improvement of signal-to-reverberation ratio by suppressing bubble waves, while not affecting the information of ocean structures embedded in the signals. This study gains insights into features of water-column acoustic reflections and provides better tools for acoustic imaging of mesoscale ocean structures.